4
\$\begingroup\$

I purchased an X-Rite ColorChecker to attempt to get better white balance for my camera (a7R IV). My first attempt at using it has yielded wildly questionable results. Process below. All files referenced can be downloaded in the zip archive here.

  1. Shoot image image.ARW
  2. Use ACR to convert to image.DNG

  3. Use X-Rite's "ColorChecker Camera Calibration" to create profile Sony ILCE-7RM4.dcp from the DNG

  4. Open original ARW or DNG in ACR.

  5. Change from default profile "Adobe Color" to the custom one I just created.

Results... beyond horrible. See below (Adobe on left, new custom profile on right). What am I doing wrong?

image

Improve this question
\$\endgroup\$
3
  • 1
    \$\begingroup\$ I can repeat your results, but idk how to fix it, or what went wrong in the first place. \$\endgroup\$
    – Tetsujin
    Jun 7, 2020 at 10:15
  • \$\begingroup\$ After further tests. If I repeat your method, as written, from my own camera & Color-Checker, I get the expected result. All I can guess [I'm on Nikon so I have no absolute comparison] is that there's "something" in your original RAW that's upsetting things. I can see nothing wrong with your actual working method. [I know, "something" really isn't helpful, sorry] \$\endgroup\$
    – Tetsujin
    Jun 7, 2020 at 11:25
  • 1
    \$\begingroup\$ I've gotten the same result using your files. It seems to be screwing with the blue calibration, almost like it was trying to null out the blues. Weird. Do you have the same problem with other calibration images, or is this a one-off problem? \$\endgroup\$
    – BobT
    Jun 7, 2020 at 15:07

1 Answer 1

Reset to default
2
\$\begingroup\$

Move the crop points slightly so that the squares are better centered in the patches before creating the profile. The problem seems to be associated with the X-Rite color checker autodetection. Even though the squares visually appear to be located within the color patches, part of the border is likely being used to create the profile.

RawTherapee developers provide guidance on shooting color calibration targets to create camera profiles. Notably, do not "fill the frame as much as possible", as some here have suggested. Rather, it's better to:

Position the target so that it fills the center-third of your frame - not more, not less. The center of the frame has the best optics and lowest vignetting.

The DCamProf developer concurs:

Uneven lighting is a common problem in camera profiling. The typical recommendation is to make sure you have even lighting (at least two lights if not shooting outdoors) and shoot the target small in the center (to minimize vignetting).

You can also try other software:

Here's the output from RawTherapee with the standard DCP tone curve and the following settings:

  1. No profile.
  2. Camera Standard.
  3. DCP file after adjusting crop points.
  4. DCP file without moving crop points (autodetect).

output

Improve this answer
\$\endgroup\$
11
  • \$\begingroup\$ Thank you, this appear to be correct. As a software engineer I've been frustrated with a number of things about the X-Rite software from the moment I first installed it. Add this to the list. Manually centering the squares does appear to yield mostly accurate results. It's annoying though that repeating that process twice does not yield identical results (though they are pretty damn close). $100 for a piece of plastic and some really poorly made software. Aggravating, but hey it's the industry standard. Yay. Thanks for solving the mystery! \$\endgroup\$
    – The111
    Jun 8, 2020 at 4:33
  • 1
    \$\begingroup\$ Yup, that makes sense. It's just surprising they can't do a better job of automatically fine-tuning the color square selection (especially since they're able to find the rough area). No worries, still probably the best offering out there, and far better than nothing. Thanks for the other link, might check that out too. \$\endgroup\$
    – The111
    Jun 8, 2020 at 6:40
  • 1
    \$\begingroup\$ Based on this answer, if you want better, more repeatable results, move over to the passport & just take a picture of it alone. That's what I did with my repro test in comments above & what I do on location too. Get your model [or chair] to hold the passport where you're measuring your light, but just zoom or walk in so there's more passport & less margin for error. \$\endgroup\$
    – Tetsujin
    Jun 8, 2020 at 13:57
  • 1
    \$\begingroup\$ I had another go at the original image, manually moving the corners. First 6 or 8 attempts the Checker software complained the sample section was too small, reinforcing my point above. The one it did manage to save was better but still not right. Take close-ups in future. \$\endgroup\$
    – Tetsujin
    Jun 8, 2020 at 14:07
  • 1
    \$\begingroup\$ @Tetsujin Yeah I learned quickly to take larger photos of the passport. Interestingly, I also found a place where the Checker software beats Adobe. I was shooting with heavy ND this weekend, like 30sec exposures. When I held the passport up for large shot, I wanted a quick handheld shot, so I jacked ISO up to max and exposure up to 1/13 sec. The result was (predictably) a blurry and grainy passport. Checker picks it up fine (auto detect), with surprisingly decent results, but Adobe refuses to process it. \$\endgroup\$
    – The111
    Jun 9, 2020 at 1:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.